Download presentation
Presentation is loading. Please wait.
Published byAnnice Haynes Modified over 5 years ago
1
This is a real picture, the physical camel is a white line on the desert floor
And hey, these camels might even be clustered!.
2
Lower 48 14064
4
California and Georgia
8
Is there a background. How many clusters are there
Is there a background? How many clusters are there? Do they have a characteristic size? Is there substructure? Are there any peculiar boundary conditions?
9
Nice approach
10
Now starting to see voids here
11
Circles are Better
13
Is most of your volume filled with Cluster or Voids
Define some void density = some under density; count in circular sizes and see how many circles of what size meet your defined threshold
15
There does seem to be a typical value
The long tail is the background – in the case where there are about as many events in the background as in the clusters, the nearest neighbor statistics will be about 1 (random) but will become much smaller if you subtract the background out. In this case there are more points in the clusters than in the background Average = 6.7 KM
16
Sliding Boxes –this approach tends to truncate the long tail
Average = 8.5 km – now you have learned something
23
Use this approach – in K-means all points are part of some cluster
High central over density compared to background
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.